Details of talk
|Title||Robust anomaly detection in non-stationary data streams|
|Presenter||Sevvandi Priyanvada Kandanaarachchi (Monash University)|
|Author(s)||Sevvandi Priyanvada Kandanaarachchi|
We present modifications to an existing anomaly detection method called HDoutliers which is a powerful algorithm based on Weissman's Spacings theorem. However, it has some limitations that affect its accuracy. In this talk, we explain Weissman's spacings theorem, and propose solutions to some limitations of HDoutliers. We also extend the modified algorithm to data streams that exhibits non-stationary behaviour.